Atmospheric Pressure vs. Low Pressure Chemical Vapor Deposition

-@§_ WATKINS-JOHNSON COMPANY l(cYolanarelic

Hee

a

oO

2

Lu a
Z WY
= =
<<

a.

(ap)

sure

18 NO

VOL.

Electrical circuitry has been evolving
to cope with ever-increasing product-
design complexity. Along with prod-
uct-design complexity has come pres-
sure to fit more and more circuit
functions in smaller spaces. Although
many of today’s technical applications
find themselves cramped for space,
electronics is the leader in achieving
functional density, as conductors and
insulators approach dimensions of a
few atoms.

Ordinary household wiring is an ex-
ample of low-tech circuitry. The tradi-
tional “wire” is a thin, round rod,
although the wire and conductor can
be of any shape. Today’s circuits,
especially complicated ones, are
printed patterns rather than complex
networks of wires.

The printing on this page is a set of
recognizable optical patterns. The
paper is an insulator on which modern
printing can produce extremely com-
plex, repeatable patterns. If the ink
were electrically conductive, it could
just as easily be a circuit pattern.

Lithography

Modern printing often utilizes a tech-
nique called lithography, which means
“stone printing.” Lithography was
invented by Alois Senefelder in 1796.
He discovered that polished stone
could be changed from ink-repelling to
ink-receptive by acid etching. By
engraving a waxed stone, he could
acid-etch selected areas and make a
durable stone printing plate. In his
case, the wax coating had to be labori-
ously hand engraved. The greater the
detail, the greater the labor. Fine detail
was, and is, fairly limited. Many other
materials have been found to respond
to this process, but the name “stone
printing” is still used. This basic tech-
nique is used in circuitry as well.

2

In the more modern process, photo-
lithography, the wax is replaced by a
photosensitive lacquer (photoresist).
The pattern artwork can be drawn on
any surface and in any convenient or
highly magnified size to handle ex-
tremely fine detail. The pattern is then
photographed and reproduced to any
scale desirable, down to the optical
resolution of the lens system. When the
image is projected on the photoresist,
an exposure pattern is produced which
is solvent-removeable. The underlying
material, be it stone or silicon wafer,
can now be subjected to selective chem-
ical etching. Patterns produced this
way can be so fine that examination
requires electron microscopy (ordinary
optical microscopy is inadequate).

Screen Printing

Another technique, silk-screen print-
ing, emerged in China around the year
500. A panel of silk was lacquered or
varnished to make it impervious to
ink. Ink could be squeezed through
unvarnished areas, however, onto an
adjacent surface to make an easily
reproduced pattern.

Printed-Circuit Boards

The January/February 1991 issue of
Tech-notes reported on methods of at-
taching circuit components to printed-
circuit boards. The “board” is a fiber-
glass sheet to which has been bonded
a thin (.003 to .005 inch) sheet of
copper. In making a printed-circuit
board, a waxy (protective) ink is screen-
printed on the copper in the desired
pattern. This is analogous to the wax-
ing of the lithography stone, but elimi-
nates the tedious engraving. The un-
covered copper is removed by acid, the
waxy ink by solvent.

Circuit Resolution

Patterns made this way typically get
no smaller than 0.020 inches (500

microns) in feature sizes. The screen-
printing scheme can be pushed a little
smaller when the ink itself is com-
pounded to be the circuit material. In
that case, the ink is printed onto a thin
ceramic plate. The ink is cured and
bonded by furnace firing. This is the
“thick-film firing” application for
which many standard furnaces are
used. The circuit pattern is produced
directly by printing. No etching or
resist removal is required. Thick-film
circuit features get down to about 0.005
inches (125 microns) in width and
.001 inch (25 microns) in thickness.

Circuit Density

Figure 1 illustrates the significant
conductor size reduction in going from
low-tech circuitry (the household wir-
ing example), to printed circuitry, to
thick-film. The idea, of course, is to
make circuits smailer or to fit more
circuitry into the same space. A wire
is round, but the conductors on PC

boards tend to be flat rectangles. Wires

tend to go from connection-to-connec-
tion in three-dimensional space,
whereas printed circuits are mostly
flat (two-dimensional). In two dimen-
sions, if the conductor and the insu-
lating spaces get ten times smaller,
the overall circuit is 100 times smaller
because both length and width are
affected. Small changes thus tend to
bring big circuit benefits. From Fig-
ure 1, one might estimate the thick-film
element to be 100 or 1000 times smaller
than household wiring, thus 10,000 to
one million times smaller for the same
circuit.

Thin Film

As dramatic as thick-film circuit reso-
lution sounds, it doesn’t stop there.
Beyond printed thick film, conductors
can be made of thin films which are
about one fiftieth of one thousandth of
an inch thick (one half micron) and

are typically about one micron wide.
This is a barely detectable speck on the
scale of Figure 1. Even in the regime of
thin film, the pressure is on to achieve
ever smaller dimensions.

ORDINARY
HOUSEHOLD WIRE
(NO INSULATION)

PC BOARD CIRCUIT
ELEMENT

APPROX. 300X

Circuitry vs. the Brain

In today’s advanced stage of technol-
ogy, microcircuits are being compared
to nerve fibers. It’s generally accepted,
however, that microcircuitry has to get
ten million times more sophisticated
before it can rival the human brain.
Any one nerve fiber is about ten mic-
rons in diameter. Although microcir-
cuit features are now smaller than
nerve fibers, nervous system intercon-
nects are far more complex, plus
they’re made in three dimensions, not
two dimensions (see Figure 2).

3

Perfection

When circuit features get extremely
small, the material out of which the
circuits are made must reach a high
degree of perfection. Small defects such
as microscopic cracks, pinholes or dirt
particles would severely damage the
circuit. This is the reason such extreme
cleanliness measures must be taken in
semiconductor manufacturing. It is
also the reason for constant research
and development to get the coatings
ever closer to absolute microscopic
perfection. As circuits become ever
smaller, one wonders, “How small is
small?” and “What is the ultimate
surface?”

In Figure 3 (Courtesy of IBM), IBM
scientists positioned individual xenon
atoms on a nickel crystal surface.
Rows of nickel atoms making up the
crystal are faintly evident. Magnifica-

4

tion is about six million. Clearly, a sur-
face is not the hard, smooth boundary
it appears to be. It is a transition zone
of atomic nebulosities. IBM went to
great extremes in achieving such indi-
vidual control of atoms. Typically, the
atomic scale makeup of any surface is
unknown. Yet, this is the region where
many vital phenomena take place. In
CVD deposition, gaseous atoms or
molecules interact one-by-one at the
surface in building up the coating.
Although microcircuits are not this
small, dimensional control of ten
atoms or so is not uncommon.

Vacuum Deposition of
Thin Films

Traditionally, thin films for microcir-
cuits are made in vacuum chambers
by various evaporation techniques.
Evaporation renders the film-forming

material gaseous. On the surface to be
coated, atom-by-atom condensation of
this evaporant forms the thin film.
The vacuum deposition process tends
to be expensive and cumbersome in
production, but more conventional
techniques (such as mechanical roll-
ing, electroplating, or paint-like coat-
ing) are totally inadequate when ex-
amined microscopically. The cost and
production inadequacies of the
vacuum process have, thus, been
largely accepted. If the highest pos-
sible control is mandatory, it would
seem that vacuum is also mandatory.

Advantages of Vacuum

Part of the logical appeal of vacuum
is that nothing could be cleaner than
absolute emptiness. Also, if air itself,
or the moisture or pollutants it con-
tains is a contaminant, then vacuum is
necessary to get rid of it. In addition,

one of the greatest sources of contami-
nation in clean rooms is people—the
workers. Flakes of skin, organic
vapors, fibers, bits of hair, and breath
and smoke residues are all attributed
to people. Since people can’t co-exist
in the vacuum chamber, vacuum pro-
vides protection from people.

Limitations of Vacuum

Where extreme control is necessary,
the merits of vacuum seem so strong
that any other method would be an
undesirable compromise. However, a
closer look shows that there are prob-
lems, even serious problems, with
vacuum and that there are more
advantages than expected with atmo-
spheric pressure.

First of all, a vacuum is never “empty.”
Even an exotic vacuum can’t remove
all the gases. Even outer space isn’t

5

totally empty. Vacuum as a pressure
concept is, thus, never zero but is some
fraction of atmospheric pressure. It
makes sense that while one might
achieve a feeling of accomplishment
in generating a high vacuum, one has
to be concerned about what’s left, not
about what’s gone. At a billionth of an
atmosphere (respectable but not exotic)
there’s still 30 billion molecules flying
around in every cubic centimeter.
Every surface atom on a super-clean
silicon wafer would experience colli-
sion or chemical reaction with one of
these molecules each second. This is
the way surface contamination occurs.
While it’s true that the same kind of
contamination would occur at atmo-
spheric pressure a billion times faster,
it’s little consolation that by resorting
to vacuum, contamination has been
avoided for a mere second.

Limitations of Vacuum Materials

Why can’t a vacuum system be truly
empty? Why can’t the pressure be truly
zero? First of all, everything (in or out
of vacuum) has a tendency to evapo-
rate without needing to liquify; like
mothballs or dry ice (sublimination).
Nothing could serve as a container for
a perfect vacuum because the container
itself vaporizes to contribute to the
residual gases. Similarly, no materials
are absolutely impermeable. Gases
from the outside slowly seep (diffuse)
through the walls of any vacuum
enclosure.

For well-chosen and well-processed
materials, sublimation and diffusion
effects can be kept small. A bigger
limitation comes from the fact that
no materials are pure or ideal. All
materials contain volatile components
to some degree and all materials
adsorb gases on their surfaces. These
volatiles are gradually liberated under
vacuum conditions, frustrating the

6

work of the pumps. The harder you try,
the more difficult it tends to get, be
cause the lower the pressure (the
higher the vacuum) the smaller the
amount of gas required to oppose your
efforts.

Gases

All gases consist of freely moving
atoms or molecules. Their number in
any given space is the real meaning of
pressure, and their velocity is the real
meaning of temperature. So any “gas”
is composed of tiny specks shooting
through a perfect vacuum at rifle-bullet
speeds. Since there are so many others
around, these specks don’t go very far
before they collide with (and bounce
off) another one. This distance is the
mean free path (MFP). At atmospheric
pressure, MFP is about one millionth
of an inch. At one tenth atmospheric
pressure, MFP is ten times longer (in-
versely proportional to pressure). Be-
cause of all the collisions, a gas mole-
cule doesn’t get from point A to B at
rifle-bullet speeds. Its random fractured
rebound motion from point A to B is
its diffusion, which at atmospheric
pressure is about one inch per second.
At one tenth atmospheric pressure,
diffusion goes to ten inches per second
(proportional to MFP).

Diffusion occurs even in gases that
are perfectly still. If there is a pres-
sure gradient, the whole body of gas
will move in response to it. Diffusion
is totally independent of bulk flow,
even moving up stream. Once pressure
flow exceeds diffusion flow, however,
no net upstream diffusion flow can
occur. This, on an atomic scale, is
what is going on when a container is
purged or flushed with clean gas. Con-
versely, flushing action cannot be
accomplished unless diffusion velocity
is exceeded.

In some cases, the object of vacuum is

to avoid diffusion collisions. Vacuum
electron tubes, nuclear particle accel-
erators, electron microscopes and ther-
mal insulation are examples. MFP
under such conditions is very long and
diffusion rate approaches rifle-bullet
speeds.Thin-film coatings, however,
are not vitally dependent on long
MFP. Are there any aspects of vacuum
inherently beneficial to coatings or is
it simply a matter of contamination
control?

Shadowing

Objects to be coated often have com-
plex non-flat surfaces. Since vacuum
deposition tends to be line-of-sight (a
consequence of long MFP), some sur-
face areas can be poorly coated if
shadowed by others. Planetary fixtur-
ing, or other means of randomizing
deposition, can help, but shadowing is
a troublesome characteristic. As cir-
cuits get more complex, the problem

gets worse. The viscous flow associated.

with higher pressure (shorter MFP)
reduces shadowing because flow under
these conditions carries material
around corners.

Radiation Damage

Sputtering processes (evaporation in-
duced by ion bombardment) depend on
complex electronics which produce
complex electromagnetic fields. The
ion and electron activity risks damage
(by bombardment) to semiconductor
substrates. This, too, is a more serious
problem with newer microcircuit struc-
tures.

Gases and Contamination

When it comes to contamination, some
gases are more objectionable than
others; some gases are totally inert and
still others are beneficial. Most of the
time, concern centers around water
vapor, carbon dioxide, carbon monox-

ide and organics as common, difficult-
to-avoid contaminants. In ordinary air,
the 80% nitrogen is often inert but the
20% oxygen is the problem.

When typical residual vacuum gases
are analyzed (see Figure 4), they are
found to contain about 70% objection-
able species. They are the result of
outgassing and degassing and are dif-
ficult to eliminate. Deposition in
vacuum tends to be slow, as the rate is
a function of pressure. This gives con-
taminants more time to do their dam-
age. Furthermore, because of long MFP
and high diffusion rate, any contami-
nation source in the whole vacuum
system tends to have easy access to
the deposition area.

It is a credit to the problem-solving
accomplishments of those working in
vacuum technology over the past 100
years that, in spite of so many limita-
tions, vacuum processes are so widely
used. Vacuum, however, is more com-
plicated than atmospheric pressure.
Atmospheric pressure processing isn’t
necessarily simple, but pressures below
(or above) atmospheric are less simple
and, thus, unfavorably impact cost
and productivity. A powerful precedent
is set in that the brain, the ultimate
in microcircuitry, requires no vacuum.
A complex product doesn’t necessarily
require more complex processing.

CVD

In the 1940’s another deposition
method began to appear. One of the
early workers with this method was
John Blocher, now retired from Bat-
telle Memorial Institute. Gaseous
sources of deposition material are used.
By comparison, in vacuum deposition,
the source material is solid, rendered
gaseous by heat or ion bombardment.
Gaseous sources can be more highly
purified than solid targets. There is
also the opportunity, using gaseous

(f

se
5 :
ped
Oy
°
=
=<.
a
2
iu
ce.
a
oS
=
oe
as
=
oo

sources, for greater variety of chemical
trickery in obtaining desired results.
It’s one thing to vacuum-deposit ele-
ments, for example, but another to
deposit compounds. Energy utilized in
subliming solid targets is often suffi-
cient to dissociate molecules, disrupt-
ing the basic makeup of the coating
material. “Reactive sputtering,” a depo-
sition technique designed to address
this problem, is used to get fractured
molecules back together again.

John Blocher’s technique, chemical
vapor deposition (CVD), is elegantly
simple in concept.

Heat + Chemicals = Coating

Heat is used to destroy (pyrolize)
gaseous molecules, leaving behind
solid residues as a coating.

Although simple in concept, it can still
be quite difficult to meet all the tough
specifications expected of the coating.

8

Partial Pressure

Of course, contamination is an impor-
tant CVD issue, as with any coating
technique. The chemicals must be very
pure since the coating cannot be any
purer than the materials it’s made
from. However, another concept is
emerging here. Purification means get-
ting rid of only the contaminants.
Vacuum techniques, on the other hand,
strive for purity (contamination con-
trol) by removing everything. If it were
indeed possible to remove everything,
it would also not be possible to produce
a coating since there are no ingre-
dients. One could argue that vacuum
evaporation doesn’t require gaseous
ingredients. But the purity of the solid
source cannot be enhanced by the
vacuum system no matter how low the
pressure is. It is not so much the total
pressure which matters. It is not so
much vacuum or atmospheric or above
atmospheric which matters, it’s the

partial pressure of the contaminants.
In the 1960’s, primitive atmospheric
pressure (AP) technology did not com-
pare well to advanced low pressure
(LP) technology. But that was a com-
parison of the level of development.
Partial pressure control at that time
favored LP. Whether AP or LP, sophis-
ticated results require sophisticated
equipment. Whether AP or LP, it
comes down to doing the job properly
to get the desired results.

Containment

It is expected that any complex process
should be carried out in an enclosure
so that process conditions can be con-
trolled and maintained. If the process
can be carried out at atmospheric pres-
sure, one important advantage, espe-
cially for production, is that work can
be brought comparatively easily in and
out of the enclosure. It would seem that
every effort should be made to process

at atmospheric pressure before resort- |

ing to pressures above or below atmo-
spheric. If desired results can be
achieved at atmospheric pressure,
what is the point of higher or lower
pressure? Developments at Watkins-
Johnson Company have shown that
desired results can indeed be achieved
at atmospheric pressure, and that
these methods work effectively in pro-
duction.

LP and PECVD

CVD is a technique which works at
any pressure, not just AP. Low pres-
sure (LPCVD) and plasma-enhanced
low pressure (PELPCVD) are also com-
monly used. A disadvantage of LPCVD
is that chemical debris must be swal-
lowed by the necessarily sophisticated
vacuum pumps. It is a remarkable
accomplishment that pumps are avail-
able which tolerate such abuse. It is
widely believed that low pressure may

have an advantage in particle forma-
tion and contamination. However, the
reality seems to be that vacuum con-
ditions, where no chemical source par-
ticles form, is a condition where no
coating forms either. Experience has
shown that particle control at AP is
quite competitve with LP.

Early APCVD for Electronics

The electronics industry experienced
great expansion in the ’50s and ’60s.
Defense application fueled much of
this expansion and was a key factor
in the formation of Watkins-Johnson
Company. The conveyor belt furnace
emerged in the mid ’60s to meet the
growing requirement for more produc-
tion-oriented equipment, primarily for
printed-circuit board and _ thick-film
applications (the big volume, low-cost
part of the industry). Also in the 1960’s,
much work was being done (by other
companies) to perfect APCVD appara-
tus and processes for electronic appli-
cations. This work was largely unsuc-
cessful and left many people with the
impression that atmospheric pressure
simply was not adequate for the task.

Three unsolved problems seemed to be
at the heart of this failure:

1. A chemical delivery nozzle suffi-
ciently accurate to form accurate
coatings was needed.

2. An enclosure design which could
protect the process area from work-
room disturbances (air contamina-
tion, drafts, etc.) was necessary.

3. A process for accurate removal of
chemical by-products was required.
(Too fast removal wastes chemical
and thins the coating; too slow,
fouls the process.)

Problem 1 caused a shift towards low
pressure. Rapid dispersion of gases

9

(when injected into a vacuum) was
used as a means of improving unifor-
mity.

Problem 2 caused a shift towards a
hermetic enclosure to seal out room
disturbances.

Problem 3 was resolved by simulta-
neously controlling gas inlet flow and
vacuum pumping speed.

The result, LPCVD, was more a failure
to solve the above three problems at
atmospheric pressure than a recogni-
tion of something inherently superior
about LP.

Early CVD at W-J

In the early 1970’s, Watkins-Johnson
Company began looking at combining
CVD with its conveyor furnace. This
work began at the urging of early
digital display makers who needed the
cost effective production-oriented fea-
tures of conveyor furnaces for their
manufacturing requirements. There
were only a half dozen or so conveyor
furnace makers in the whole world
they could turn to. The furnace em-
bodies a significant amount of tech-
nology in itself. CVD is also a signifi-
cant technology by itself. Developing
both, simultaneously, could be over-
whelmingly difficult, but Watkins-
Johnson Company soon developed
significant CVD proficiency.

The three problems, listed above, were
solved by:

1. Development of a unique nozzle
(injector) with performance features
which reach semiconductor specifica-
tions. The injector is totally manu-
factured at Watkins-Johnson, using
equipment and techniques developed
at Watkins-Johnson.

2. Experience with elaborate furnace
muffle structures provided a launch
point for still greater sophistica-

10

tion. CVD muffles can now provide
the required process protection and
work-room isolation without having
to be hermetic.

3. A unique self-cleaning vent-line
flow meter has been developed
which provides controlled by-
product removal.

CVD Evolution at W-J

Early digital display coating require-
ments were quite a bit less stringent
than those for semiconductors. The
first Watkins-JJohnson Company CVD
furnace was sold in 1975 and it is still
in display-making service. Overall
proficiency improved to the point
where in the early 1980’s semiconduc-
tor applications could be considered.
The conveyor furnace, originally for
thick film, had now become thin-film
qualified.

A major milestone in the mid 1980’s
was the successful demonstration of
borophospho-silicate glass (BPSG)
coatings. No low-pressure process has
yet been able to demonstrate full pro-
duction practicality. For the first time,
an atmospheric pressure process was
(at least for BPSG) superior to
vacuum. [3].

LP Limitations

The BPSG achievement brought fur-
ther awareness to inherent limitations
of low-pressure techniques. Batch
processing was necessary at low pres-
sure because low LP coating rates
could not otherwise keep up with pro-
duction. While 100 wafers at a time
was Okay for smaller wafers and sim-
pler circuits, it’s not okay today. The
problem is that the same coating con-
ditions can not be delivered to each of
the batched wafers. Gas phase reaction
plus coating depletion rapidly changes
the gas composition as it flows over
the wafers. In addition, larger wafers

disturb flow patterns and mask each
other more severely than small ones.
Another factor is that complex coat-
ings, especially from organic precur-
sors, need large oxygen excess to react
properly. In a vacuum system, exces-
sive oxygen results in depression of
deposition chemicals, causing lower
deposition rate, further aggravating an
already-low deposition rate problem.

Single Wafer

These developments shook the LP
industry to where batch processes have
now all been abandoned in favor of
single wafer. Conveyorized APCVD is
essentially single wafer in the sense
that all wafers get the same treatment.
It’s obtained quite naturally at AP
while it’s quite a struggle at LP. LP
processes have been gradually rising
in pressure to increase deposition rate.
Some approaching atmospheric pres-
sure are called “high pressure” or “sub-
atmospheric.” |

Non-Oxide Coatings

One advantage to vacuum processing
had been in the preparation of oxygen-
sensitive films. Metal or silicon films,
for example, would be severely conta-
minated by air getting into the open
ends of conveyorized CVD. For a while
it was thought that APCVD might
only be suitable for oxides; all others
would need LP. However, development
during the 80s has shown this isn’t
so[4]. The reason gets back to the
nature of diffusion. It was pointed out
earlier that coating processes, vacuum
or otherwise, are crucially dependent
on source material purity. If a shower
of similarly pure inert gas (carrier gas)
covers the process area, outside air (or
any other contaminant) can not get in,
provided the flow velocity exceeds dif-
fusion velocity (see Figure 5). Since dif-
fusion velocity is only one inch per

second at atmospheric pressure,
modest flows are sufficient to provide
protection. The belt and its load of
wafers go in the muffle, but air doesn’t.
This is quite a nice atmospheric advan-
tage. Anything comparable at LP
requires complicated interlocks. Diffu-
sion works for you at AP but against
you at LP. This is the reason oil back-
streaming is a common vacuum prob-
lem. There are non-oily pumps but the
oily ones tend to be the good work
horses.

Purity

There has certainly been increased
attention given to gas and chemical
purity in recent years. The availability,
in large production, of high-purity cyro-
genic nitrogen has been an impressive
industrial feat. The emergence of chem-
ical getter point-of-use style gas polish-
ing filters has been another impressive
development. What is happening is
that the reduction in partial pressure
of impurities is progressing faster than
the ability of vacuum systems to
reduce total pressure. Some chemical
getter filters can now reduce oxidizing
impurities in nitrogen to levels lower
than the best ultra-high vacuum sys-
tems. While exotically cleaned piping
is mandatory just as in ultra-high
vacuum, the role of low total pressure
(vs. low partial pressure) is becoming
uncertain except where long mean-free-
path is required.

Low-Temperature Deposition

An important advantage of plasma-
enhanced low-pressure CVD (and to
some extent photo-enhanced CVD) is
the ability to induce low-temperature
deposition. APCVD, which is simply
thermally induced, requires, for ex-
ample, about 650°C to make silicon
nitride, whereas plasma works at
about 250°C. There are some qualifica-
tions, however. Plasma nitride, al-

11

though it is useful material, is not
really silicon nitride. The proportions
of silicon and nitrogen vary; oxygen is
often present, and hydrogen is almost
invariably present. The hydrogen may
be given off in subsequent processing,
causing damage to the microcircuit
structure. Plasma processing tends to
be slow, have particle problems and, of
course, risk radiation (ion or electron
bombardment) damage. Still, when
process temperature is a crucial factor,
plasma has a clear advantage.

Atmospheric Plasma

Plasma doesn’t have to be a low-pres-
sure process. It is simply a matter of
coincidence that a plasma is easier to
generate at one thousandth of an atmo-
sphere or so. Free electron mean-free-
path at this pressure provides an opti-
mum for easy ignition. Lasers, x-rays,
nuclear radiation, corona discharges
and sparks can produce plasmas at
atmospheric pressure, but relatively
little work has been done to associate
them with deposition. One of the new

12

deposition techniques depends on
corona discharge at atmospheric pres-
sure to generate point-of-use ozone.
Deposition is radically altered towards
lower temperature by ozone vs. oxygen.
The point is that plasma and AP are
not mutually exclusive. Ozone is one
example of atmospheric plasma and is
likely to lead to others. There has also
been some chemical solution to deposi-
tion temperature reduction.

Precursors

In the 1950’s, when microcircuit tech-
nology was developing, the usual pre-
cursor for silicon dioxide films was
ethyl silicate (more correctly, tetra
ethyl ortho silicate; TEOS). The depo-
sition temperature, 800°C or so, was
undesirably high, prompting a search
for a better-behaved chemical. Silane
was considered the answer. A long and
substantial silane-based industry re-
sulted. Interestingly, because of its
superior conformality on cramped
modern circuitry, plus temperature
reduction brought by ozone, silane is

now being upstaged by TEOS. Another
intriguing example is the development
at Harvard University of titanium nit-
ride precursors which work at 200°C
and at atmospheric pressure. Precur-
sors for low temperature AP, silicon
nitride may not be far behind.

Liquid Source

TEOS ozone and BPSG from liquid
sources (reacted with ozone) is the
object of Watkins-Johnson Company’s
newest CVD furnace design. Intense
equipment design and process devel-
opment has occurred, and with marked
success, in response to the industry’s
urgent need for improved conformality
and reduced process temperature [5}.

Conclusions

The TEOS ozone CVD furnace is but
the newest of a long series of develop-

ments spanning the whole history of

Watkins-Johnson Company. Furnaces
began in the early 1950’s as laboratory
specials for Stanford University by
Ray Stewart (Founder of Stewart
Engineering which became W-J
Stewart Division in 1964). He used a
primitive form of CVD as early as 1960
(charge dissipating coating of tin
oxide, inside backward wave oscillator
glass vacuum envelopes). Conveyor
furnaces emerged in the mid-sixties
and conveyorized CVD in the early
"10's.

Today, CVD is primarily focused
on electronics. But that’s just the begin-
ning. Substantial growth potential
exists. Without the restrictions of
vacuum, heavy industry can also be
served. Watkins-Johnson Company
already interacts in small ways in flat
panel displays, solar energy, architec-
tural glass, lighting and protective
coatings. There are applications in
smart mirrors, smart windows, auto-

motive windshields, fiber optics, copy-
ing machines, thin-film electronic
devices, new types of thin-film cir-
cuitry, coatings on plastic, wear, abra-
sion, corrosion resistant coatings and
so on. So many applications arise from
the fact that coatings always upgrade
the performance of a substrate in some
way. The world we live in is a vast
array of surfaces which could all stand
to be improved in some way. Truly, we
have barely scratched these surfaces.

References

1. Gralenski, N. “Thin Films By Conveyor-
ized Atmospheric CVD,” ISHM—
Internepcon, Tokyo, Japan, 1983.

2. Gralenski, N. “Advanced APCVD
Reactors for Thin Film Deposition,”
Microelectronic Manufacturing and
Testing, Sept., Oct., 1987.

3. Gralenski, N. “Advanced APCVD:
BPSG Film Quality and Production
Reliability Report,” Semicon West,
San Mateo, Ca., May 21, 1986.

4. Bartholomew, L.D. and S.M.
McDaniel. “Hydrogen Reduced CVD
Blanket Tungsten Deposited at
Atmospheric Pressure,” MRS Work-
shop on Tungsten and Other
Advanced Metals VII, October 22,
1990.

5. Bartholomew, L.D. and J.C. Sisson.
“Optimization of APCVD TEOS/O3
Process for SiOg and BPSG,” Schu-
macher Dielectrics and CVD Metal-
lization Symposium, February 11,
1991.

13

Author:

Nicholas Gralenski

Mr. Gralenski is Staff Scientist and
Manager of New Technology Develop-
ment at the WatkinsJohnson Com-
pany Scotts Valley facility, and is the

developer of the conveyorized CVD
furnace.

He is involved in many of the ongoing
CVD development projects. He has
written numerous technical papers and
presented talks on the subject of CVD.
He interfaces with Marketing and
Sales to keep technical feasibility in
line with customer needs.

Mr. Gralenski is a physicist, with
strong minors in chemistry and
mechanics, obtaining his B.S. degree
from the University of Massachussetts.

14

Notes

15

WATKINS-JOHNSON COMPANY
3333 HILLVIEW AVENUE

PALO ALTO, CA 94304-1204
(415) 493-4141

ADDRESS CORRECTION REQUESTED

WATKINS-JOHNSON Facility
Locations

United States
CALIFORNIA

Watkins-Johnson

3333 Hillview Avenue
Palo Alto, 94304-1204
Telephone: (415) 493-4141

Watkins-Johnson

2525 North First Street
San Jose, 95131-1097
Telephone: (408) 435-1400

Watkins-Johnson

440 Kings Village Road
Scotts Valley, 95066-4081
Telephone: (408) 438-2100

international

UNITED KINGDOM ITALY

Watkins-Johnson
Dedworth Road

Watkins-Johnson S.p.A.
Piazza G. Marconi 25

BULK RATE
U.S. POSTAGE

SAN JOSE
CALIFORNIA

MARYLAND

Watkins-Johnson

700 Quince Orchard Road
Gaithersburg, 20878-1794
Telephone: (301) 948-7550

Watkins-Johnson

8530 Corridor Road
Savage, 20763

Telephone: (301) 497-3900

NORTH CAROLINA

Watkins-Johnson

1000 West Powell Road
Fuquay-Varina, 27526-9399
Telephone: (919) 552-6161

GERMANY,
FEDERAL REPUBLIC OF

Watkins-Johnson
Boschstr. 10

Oakley Green 00144 Roma-EUR 8039 Puchheim

Windsor, Berkshire SL4 4LH Telephone: 6 592 4554 Telephone: 89 802087/88
Telephone: 753 869241 6 591 2515 Telex: 529 401

Telex: 847 578 Telex: 612 278 Cable: WJDBM-MUENCHEN
Cable: WJUKW-WINDSOR Cable: WJ ROM | Telefax: 89 803044

Telefax: 753 841534

The Watkins-Johnson Tech-notes is a bi-monthly periodical circulated to educational institu-
tions, engineers, managers of companies or government agencies, and technicians. Individuals
may receive issues of Tech-notes by sending their subscription request on company letterhead,
stating position and nature of business to the Editor, Tech-notes, at Watkins-Johnson’s Palo
Alto, California address. Permission to reprint articles may also be obtained by writing the Editor.

COPYRIGHT © 1991 WATKINS-JOHNSON COMPANY

Printed in U.S.A.